@nanduri_anil @magicsilicon Agree. Still, assuming you agree that hyperscalers who contribute ~50% of Nvidia's revenue may switch if a decent alternative appears, the CUDA moat may not be as deep or as broad as people imagine.
View original →@nanduri_anil @magicsilicon Agree. Still, assuming you agree that hyperscalers who contribute ~50% of Nvidia's revenue may switch if a decent alternative appears, the CUDA moat may not be as deep or as broad as people imagine.
View original →@nanduri_anil @magicsilicon How about hyperscalers who seek to escape/avoid total lock-in by Nvidia? Do they also need a 10X improvement to find it worthwhile? Nvidia derives ~50% of revenue from just a handful of hyperscalers.
View original →@loic_na @magicsilicon Competitors will eventually catch up and become good enough. Perhaps not in all aspects but in specific workloads. Nvidia spending $20B to take out Groq may be evidence?
View original →@loic_na @magicsilicon I agree. My point is - for Nvidia to remain king, they must maintain leadership performance and TCO. CUDA is important in that it contributes to that, plus, as you said, time-to-market. CUDA as "ecosystem moat" is not as deep as people believe, IMO.
View original →@magicsilicon *IF* they ever fall behind in the perf/TCO dept, CUDA, purely as an ecosystem moat may not be as invincible as people think?
View original →@magicsilicon I guess what I'm trying to say is the real reason Nvidia remains King is that they are driving harder and faster than ever, and continue to deliver the most performant, lowest TCO solutions out there.
View original →@magicsilicon Nvidia derives almost 50% of its revenue from a handful of hyperscalers, all of whom have the resources/ability and are eager to escape lock-in. OTOH, x86 customers are numerous and diverse, most of whom don't have strong incentives (or ability) to switch.
View original →@magicsilicon Thanks for writing this. I have a question: The CUDA moat may be deeper than x86, but is it unbreakable? Suppose someone comes up with a solution that materially outperforms Nvidia GPUs in real-world performance and TCO, do you believe Nvidia's moat will not weaken?
View original →@gindi4711 1. They get x86. CPUs are proving important for AI. Jensen himself acknowledged the relevance and longevity of x86, investing $5B and co-developing custom x86 CPUs with Intel. 2. Arm is now a competitor. 3. One advantage AMD has over Nvidia is they have both GPUs and x86 CPUs.
View original →@type1ayy Curious why you would vote against it, as an Nvidia shareholder. You don't think buying Intel would benefit Nvidia?
View original →